Comparison of the MOVES2010a, MOBILE6.2, and EMFAC2007 mobile source emission models with on-road traffic tunnel and remote sensing measurements

Abstract

The Desert Research Institute conducted an on-road mobile source emission study at a traffic tunnel in Van Nuys, California, in August 2010 to measure fleet-averaged, fuel-based emission factors. The study also included remote sensing device (RSD) measurements by the University of Denver of 13,000 vehicles near the tunnel. The tunnel and RSD fleet-averaged emission factors were compared in blind fashion with the corresponding modeled factors calculated by ENVIRON International Corporation using U.S. Environmental Protection Agency's (EPA's) MOVES2010a (Motor Vehicle Emissions Simulator) and MOBILE6.2 mobile source emission models, and California Air Resources Board's (CARB's) EMFAC2007 (EMission FACtors) emission model. With some exceptions, the fleet-averaged tunnel, RSD, and modeled carbon monoxide (CO) and oxide of nitrogen (NO_x) emission factors were in reasonable agreement (±25%). The nonmethane hydrocarbon (NMHC) emission factors (specifically the running evaporative emissions) predicted by MOVES were insensitive to ambient temperature as compared with the tunnel measurements and the MOBILE- and EMFAC-predicted emission factors, resulting in underestimation of the measured NMHC/NO_x ratios at higher ambient temperatures. Although predicted NMHC/NO_x ratios are in good agreement with the measured ratios during cooler sampling periods, the measured NMHC/NO_x ratios are 3.1, 1.7, and 1.4 times higher than those predicted by the MOVES, MOBILE, and EMFAC models, respectively, during high-temperature periods. Although the MOVES NO_x emission factors were generally higher than the measured factors, most differences were not significant considering the variations in the modeled factors using alternative vehicle operating cycles to represent the driving conditions in the tunnel. The three models predicted large differences in NO_x and particle emissions and in the relative contributions of diesel and gasoline vehicles to total NO_x and particulate carbon (TC) emissions in the tunnel.

Implications:

Although advances have been made to mobile source emission models over the past two decades, the evidence that mobile source emissions of carbon monoxide and hydrocarbons in urban areas were underestimated by as much as a factor of 2–3 in past inventories underscores the need for on-going verification of emission inventories. Results suggest that there is an overall increase in motor vehicle NMHC emissions on hot days that is not fully accounted for by the emission models. Hot temperatures and concomitant higher ratios of NMHC emissions relative to NO_x both contribute to more rapid and efficient formation of ozone. Also, the ability of EPA's MOVES model to simulate varying vehicle operating modes places increased importance on the choice of operating modes to evaluate project-level emissions.

Introduction

On March 2, 2010, the U.S. Environmental Protection Agency (EPA) announced the official release of the Motor Vehicle Emissions Simulator (MOVES2010) model for use in state implementation air quality plan (SIP) submissions to EPA and regional emission analysis for transportation conformity determination outside of California (Federal Register, 2010). MOVES2010 is the latest upgrade to EPA's modeling tools for estimating emissions from cars, trucks, motorcycles, and buses, and it replaces the MOBILE6.2 model (EPA, 2010). It was developed by EPA, in part, as a response to a National Research Council (NRC) review of the MOBILE model (NRC, 2000). The NRC found that although MOBILE is suited for aggregate regional and national analyses of emissions and air quality, it could not be used for assessment of mobile source emissions at temporal and spatial scales relevant to specific transportation projects and control measures. Exhaust emission factors from the MOBILE model and the California Air Resources Board's (CARB's) EMission FACtors (EMFAC) model are based upon cycle-average emissions that are corrected for average speed. The MOVES model provides greater flexibility to evaluate project-level emissions by allowing the user to input any vehicle operation cycle and estimate running exhaust emissions as a function of vehicle-specific power, the instantaneous power demand of the vehicle divided by its mass. The Federal Register notice marked the beginning of a 2-yr transition phase for MOVES, which was recently extended for transportation conformity analysis to March 2, 2013 (EPA, 2012).

An important finding of the NRC review of the MOBILE model in the late 1990s was that EPA had not addressed model validation and evaluation adequately during development of the model and recommended evaluation studies involving field observations (e.g., ambient air measurements, tunnel studies, and remote sensing), air quality modeling, and vehicle emission data (e.g., data from vehicle emission inspection and maintenance programs, roadside pullover inspections, and other direct tailpipe emission measurements) to validate model outputs. These recommendations are also applicable to the MOVES model and the FACA (Federal Advisory Committee Act) MOVES Workgroup, which was formed to provide input on various issues regarding the development of the MOVES model, recommended that EPA give high priority to validation of the new model (Barth, 2009). Although advances have been made to mobile source emission models over the past two decades, the evidence that mobile source emissions of carbon monoxide and hydrocarbons in urban areas were underestimated by as much as a factor of 2–3 in past inventories underscores the need for on-going verification of emission inventories (Chico et al., 1993; Fujita et al., 1992; Harley et al., 1993; Ingall et al., 1989; Pierson et al., 1990; Wagner and Wheeler, 1993;).

In this study, Desert Research Institute (DRI) conducted an on-road mobile source emission study at the traffic tunnel on Sherman Way in Van Nuys, California (Van Nuys tunnel), in August 2010 to measure fleet-averaged fuel-based emission factors of regulated and unregulated pollutants. The measured emission factors were compared with corresponding fleet-averaged emission factors estimated using MOVES2010a, MOBILE6.2, and EMFAC2007. ENVIRON International Corporation (ENVIRON) provided the vehicle-specific emission factors for ambient temperature and traffic conditions observed during the Tunnel Study. We also compared the average of emission factors measured for about 13,000 vehicles by the University of Denver with a remote sensing device (RSD) with the average of the corresponding modeled emission factors. The RSD measurements were made on August 12–16, 2010, a week prior to the start of the Tunnel Study on eastbound Sherman Way about 600 m west of the Van Nuys tunnel (Bishop et al., 2012). Data from each of the three study participants were submitted in a blind manner to one of the study's coauthors (D. Lawson) at the National Renewable Energy Laboratory, so that they could be independently evaluated before they were shared among all study participants.

Experimental Methods

On-road measurements

The field study at the Van Nuys tunnel was completed during a 2-week period beginning August 17, 2010. Measurements and samples were obtained during two 3-hr sampling periods each day (∼09:00 a.m. to 12:00 noon and 12:15 p.m. to 15:15 p.m.) on August 20, 21, 22, 24, 25, 26, 28, and 29 (two Saturdays, two Sundays, and four weekdays). The tunnel is located on Sherman Way, a major east-west arterial street with three lanes of traffic in each direction, where it passes under the runway of the Van Nuys Airport. In-tunnel measurements were made in the eastbound bore. The eastbound roadway grades from the west to east portals are 1.7% downgrade from 0 to 29.5 m, 0.4% downgrade from 29.5 to 111.5 m, 0.3% upgrade from 111.5 to 205.1 m, and 1.0% from 205.1 to 239 m. The distance-weighted average grade is 0.088% downgrade. The traffic directions are separated by a concrete wall, with eight door-size openings in the dividing wall, spaced evenly through the tunnel. The eight open doorways account for about 0.3% of the interior surface area of one bore of the tunnel. The tunnel is 239 m long, portal-to-portal, with a traffic turnout in the eastbound bore of the tunnel 147 m from the traffic entrance and 75 m from the exit of the tunnel. The tunnel is sufficiently far from local neighborhood housing that vehicles passing through the tunnel are being driven under hot, stabilized operating conditions.

Monitoring equipment and volatile organic compound (VOC) samplers were operated inside a minivan parked in the turnout and particle samplers were located in a small pull trailer attached to the van. Power for the samplers was supplied by a diesel generator located on the airport runway apron above the eastbound tunnel exit. Parallel measurements and samples were also obtained on airport property above the west edge of the tunnel to measure the urban background pollutant concentrations already present in the air entering the tunnel. A video camera at this location recorded the traffic entering the tunnel. An on-board diagnostics (OBDII)-based vehicle data logger was installed on a car that was driven through the tunnel up to 10 times during each 3-hr sampling period. The recorded vehicle performance parameters include second-by-second vehicle speed and 5-sec average engine speed, engine load, and throttle position (mean of 41 miles per hour [mph] for 69 measurements through the tunnel).

Table 1 shows the average temperature, relative humidity, and traffic volume, speed, and average vehicle counts per hour by vehicle type during each of the sampling periods for which vehicle counts and complete analytical data were obtained. Video from the first three sampling periods were unusable due to poor resolution and incomplete recordings. The video camera was replaced with a higher-resolution model during the second day of sampling between the morning (AM) and afternoon (PM) periods. Canister samples were invalid for the August 22 AM period due to a sampling error. Sample collection during the August 28 PM period was interrupted due to failure of the generator that supplied power to the in-tunnel samplers. Time-integrated samples were collected on August 29 for gas-phase pollutants only, using battery power for the in-tunnel samplers.

Table 1. Temperature, relative humidity, traffic volume, and vehicle speeds for sampling periods with complete valid data

Vehicles Per Hour
Date	Day or Type	Time	Temperature (°F)	RH (%)	Speed (mph)	All Vehicles	All Diesels	HDV (>10,000 lbs GVW)
August 21	Sat	PM	95	16	41.4	1486	37	25
August 22	Sun	PM	92	22	45.0	1290	8	7
August 24	Tue	AM	92	19	40.9	1335	59	25
August 24	Tue	PM	101	14	41.9	1654	64	27
August 25	Wed	AM	92	21	40.1	1419	57	10
August 25	Wed	PM	102	19	41.6	1335	59	13
August 28	Sat	AM	72	54	45.9	1294	26	8
August 29	Sun	AM	70	53	42.3	1048	13	9
Average	Weekday		97	18	41.1	1436	60	19
Average	Sat		83	35	43.7	1390	32	17
Average	Sun		81	37	43.6	1169	10	8
Average	High temp. (101–102 °F)		102	17	41.7	1495	62	20
Average	Medium temp. (92–95 °F)		93	19	41.8	1383	40	17
Average	Low temp. (70–72 °F)		71	53	44.1	1171	19	9
Average	All		89	27	42.4	1358	40	16
Note: The average vehicle speeds shown for each sampling period are based on averages of up to 10 trips through the tunnel using an OBD data logger.

After initial review of the analytical data, we eliminated three additional runs from further consideration. Measurements on August 26 were affected by road construction about 50 m west of the tunnel on the north side of Sherman Way. One lane was blocked due to the construction and traffic frequently backed into the tunnel on the westbound lanes. This resulted in higher-than-normal pollutant concentrations at the background site, which was located above the west end of the tunnel. Additionally, the potential was greater for contributions of pollutants from westbound traffic to the eastbound bore through the small door-size openings at this end of the tunnel. The Sunday, August 29, PM period was also excluded because net CO₂ was approximately equal to the analytical uncertainty, resulting in unacceptably large uncertainty in the calculated fuel-based emission factors.

The videotapes were processed by an experienced automotive mechanic, who logged the model year and EMFAC category for each vehicle entering the tunnel for a subset of weekday and weekend runs. The resulting distributions were used to estimate fleet composition for other runs for which less detailed vehicle counts were provided. For light-duty vehicles only, we used model year distributions from the RSD data set obtained by University of Denver from the California Department of Motor Vehicles (DMV) on eastbound Sherman Way during the week preceding the tunnel experiment, which are more accurate because they are based on DMV license records. Because the observed variations in fleet distributions were small from run to run, it is not expected that any inaccuracies in the extrapolated distributions would have a significant effect on model results.

Sampling and analysis methods

Measurements made during the Tunnel Study included the following: nitric oxide (NO) and oxides of nitrogen (NO_x) with a Horiba chemiluminescence analyzer (Irvine, CA) and 2B Technology 400/401 analyzer (Boulder, CO); carbon dioxide (CO₂), carbon monoxide (CO), methane (CH₄), and C₂–C₁₁ speciated hydrocarbons from 3-hr canister samples; PM_2.5 (particulate matter with aerodynamic diameter ≤2.5 μm) mass and elements from 3-hr filter samples on Teflo (2.0 μm pore size, 47 mm diameter Teflon filters [RPJ047] sampling at 56.6 liters per minute [lpm]; Ann Arbor, MI); organic carbon/elemental carbon [OC/EC] from 3-hr Pallflex 47 mm diameter prefired quartz filters (2500 QAT-UP) sampling at 56.6 lpm; C₁–C₇ carbonyl compounds collected on Waters Sep-Pak 2,4-dinitrophenylhydrazine (DNPH) cartridges (Milford, MA) sampling at 1 lpm for 3 hr; and semivolatile organic compounds (SVOCs) and PM organic speciation from Teflon-impregnated glass fiber filters (8 × 10 inch, T60A20 Pallflex; Radnor, PA) backed up by 4 inch diameter XAD cartridges collected with TE-PNY1123 Accuvol modified TIGF/XAD high-volume samplers (Tisch Environmental, Cleves, OH) operating at approximately 280 lpm for 6 hr (combined 3-hr AM and 3-hr PM sampling periods). Meteorological measurements were obtained at the background site and supplemented with airport data.

Methane, CO, and CO₂ were measured from canister samples using a Shimadzu GC-17A gas chromatograph (GC; Pleasanton, CA) with flame ionization detector (FID) and a 20 feet × 1/8 inch inner diameter (i.d.) column, packed with a 60/80 mesh of Carboxen 1000 (Supelco, St. Louis, MA). CO and CO₂ were first converted to methane by a methanator (firebrick powder impregnated with nickel catalyst) positioned between the GC column and the FID. The FID response was calibrated with the gaseous standard mixtures (Scott Specialty Gases, National Institutes of Standards and Technology [NIST] traceable; Plumsteadville, PA) containing CO, CO₂, and CH₄ in zero air. The minimum detection limit for CO, CH₄, and CO₂ were 0.06, 0.2, and ∼3 ppmv, respectively, with precision better than 10%.

Speciated C₂–C₁₁ hydrocarbon compounds were measured using gas chromatography/mass spectrometry (GC/MS) technique according to EPA Method TO-15 (EPA, 1999a). The GC-FID/MS system includes a Lotus Consulting Ultra-Trace Toxics sample preconcentration system (Lotus Consulting, Long Beach, CA) built into a Varian 3800 GC (Walnut Creek, CA) with FID coupled to a Varian Saturn 2000 ion trap MS. Light hydrocarbons (C₂–C₄) are separated on a Chrompack Al₂O₃/KCl column (25 m × 0.53 mm × 10 μm) leading to FID. The mid-range and heavier hydrocarbons (C₄–C₁₁) are deposited to a J&W DB-1 column (60 m × 0.32 mm × 1 μm) connected to the ion trap MS. The GC initial temperature is 5 °C, held for approximately 9.5 min, then ramps at 3 °C/min to 200 °C for a total run time of 80 min. Calibration of the system is conducted with a mixture that contained the most commonly found hydrocarbons (75 compounds from ethane to n-undecane; purchased from Air Environmental) in the range of 0.2–10 ppbv.

C₁–C₇ carbonyl compounds were measured as their hydrazone derivative according to EPA Method TO-11A (EPA, 1999b) using a high-performance liquid chromatograph (HPLC; Waters 2690 Alliance HPLC System with 996 Photodiode Array Detector). After sampling, the cartridges were eluted with acetonitrile. An aliquot of the eluent was transferred into a 2-mL septum vial and injected with an autosampler into a Polaris C18-A 3 μm 100 × 2.0-mm HPLC column.

The Teflon filters were weighed on a Mettler Toledo MT5 electro microbalance (Columbus, OH) and analyzed for elements by energy-dispersive X-ray fluorescence (EDXRF) analysis on a PANalytical Epsilon 5 EDXRF analyzer (Westborough, MA). PM samples were also analyzed by inductively coupled plasma mass spectrometry (ICP-MS) for total Mg, Al, Ca, V, Cr, Mn Fe, Ni, Cu, Zn, Mo, Ba, Ce, Hg, and Pb. The quartz filters were analyzed for EC and OC by thermal optical reflectance (TOR) method (Chow et al., 2001) using the IMPROVE_A (Interagency Monitoring of Protected Visual Environments) temperature/oxygen cycle protocol (Chow et al., 2007).

Semivolatile and condensed-phase organic species that were identified and quantified from the TIGF filters and XAD-4 resin samples included 55 polycyclic aromatic hydrocarbons (PAHs), 23 hopanes and steranes, and 50 alkanes and cycloalkanes in the C₁₂–C₄₀ range using a modified EPA Method TO-13A (EPA, 1999c; Wang et al., 1994a, 1994b). The higher-molecular-weight ∼C₂₀–C₃₅ alkanes and cycloalkanes in lubricating oils appear as a single hump on the gas chromatograms. These compounds were quantified together based on the ions with mass-to-charge ratios (m/z) of 57 and 55, which are characteristic aliphatic hydrocarbons, and reported as total unresolved complex mixture (UCM) of alkanes. The TIGF filters and XAD-4 resins were extracted separately using the Dionex ASE (Sunnyvale, CA) with dichloromethane followed by hexane extraction under 1500 psi at 70 °C. Prior to extraction, the following deuterated internal standards were added to each filter and XAD-4 sorbent: naphthalene-d8, biphenyl-d10, acenaphthene-d10, phenanthrene-d10, anthracene-d10, pyrene-d10, benz[a]anthracene-d12, chrysene-d12, benz[k]fluoranthene-d12, benzo[e]pyrene-d12, benzo[a]pyrene-d12, perylene-d12, benzo[g,h,i]perylene-d12, coronene-d12, cholestane-d6, hexadecane-d34, eicosane-d42, tetracosane-d50, octacosane-d58, and triacontane-d62. All extracts were concentrated by rotary evaporation at 35 °C under gentle vacuum to ∼1 mL and filtered through 0.2-μm polytetrafluoroethylene (PTFE) disposal filter (Whatman Pura disc 25TF; Florham Park, NJ), rinsing the flask three times with 1 mL dichloromethane and hexane (50/50 v/v) each time. The solvent was exchanged to toluene under ultra-high-purity nitrogen.

The TIGF filters and XAD-4 extracts were analyzed separately by GC/MS, using a Varian CP-3800 GC equipped with a CP8400 autosampler and interfaced to a Varian 4000 ion trap for analysis of all semivolatile and condensed-phase organic compounds except hopanes and steranes, as described before (Fujita et al., 2007). Hopanes and steranes were analyzed using the Varian 1200 triple quadrupole gas chromatograph/mass spectrometer (GC/MS/MS) system with CP-8400 autosampler due to the higher sensitivity of this system. Quantification of the individual compounds was obtained by the selective ion mode (SIM) technique, monitoring the molecular (or the most characteristic) ion of each compound of interest and the corresponding deuterated internal standard.

Calculation of emission rates

Composite fleet-averaged fuel-based emission rates were determined using the carbon balance method shown in eq 1 (Ban-Weiss et al., 2008; Fraser et al., 1998; Kirchstetter et al., 1999). In this method the increase of CO₂, CO, and organic gases plus organic PM within the tunnel is proportional to the amount of carbon that was present in the fuel consumed, and emission factors for each pollutant species per unit fuel consumed can be computed based on a carbon balance in the tunnel. Most of the carbon in gasoline and diesel fuel is emitted as CO₂, with smaller amounts emitted as CO. Even smaller amounts of fuel carbon emitted as PM and unburned hydrocarbons are typically left out of the denominator in eq 1. The emission factor E _P (g of pollutant P per kg fuel burned) can be calculated as

(1)

where Δ[P] is the background-subtracted (tunnel minus background) mass concentration of pollutant P (μg/m³). The fuel carbon components are similarly background-subtracted concentrations in mg C per m³. The fuel carbon mass fraction, w _c, is 0.85 g C per g fuel. This method is insensitive to uncertainties in air flow rates through the tunnel and is insensitive to the effects of any small air exchange between adjacent bores of the tunnel.

Modeled emission rates

Three motor vehicle emission models were used to compare modeled emission rates with fleet emission rates measured in the Van Nuys tunnel: (1) California EMFAC2007, (2) EPA MOBILE6.2, and (3) EPA MOVES2010a. Each model was run by ENVIRON to provide emission rates (in gram per mile units) by model year within each vehicle category for total hydrocarbons (THC), CO, NO_x, PM_2.5, and CO₂. EC and total carbon (TC) estimates were also evaluated, although only MOVES calculated these parameters directly. For MOBILE, which only provides EC and OC for diesel vehicles, GASPM (gasoline vehicle PM emissions) minus SO4 (gasoline vehicle sulfate particle emissions) was used as substitute for TC from gasoline vehicles. For EMFAC, we estimated diesel TC as PM2.5 exhaust − k × SO2, where k = 0.0694 is the slope of regression of SO₂ and sulfate for all diesel vehicles (1985–2010) in MOBILE6.2 (r ² = 0.985). For gas vehicles, we used TC = PM2.5 exhaust. Because measured background-subtracted THC in the tunnel had a high degree of uncertainty due to the relatively large background methane concentrations, the measured background-subtracted nonmethane hydrocarbon (NMHC) was used for the comparison. Modeled THC emission rates were adjusted to approximate NMHC by subtracting an average fraction of methane derived from gasoline and diesel exhaust speciation profiles (CARB, 2000).

In the EMFAC and MOBILE models, vehicle emission rates represent averages over a driving schedule with a defined average speed. The MOVES2010a model estimates emissions using relative time and emission rates in vehicle speed and specific power bins. MOVES2010a provides default test cycles to simulate the approach used for EMFAC and MOBILE6.2, or can use customized drive cycles, as was done in this study. Speed trace samples in increments of ±1 mph lasting about 9–10 sec each were collected for 69 drive-through events during the actual tunnel sampling periods with an average speed of 40.9 ± 0.8 mph. Average start and end speeds relative to the average speed were −0.2 and −0.7 mph, respectively, and the highest and lowest speeds relative to the average speed were +3.6 and −4.6 mph. Average acceleration was −0.03 m/sec² and the maximum acceleration and deceleration was +0.9 and −1.3 m/sec². The closest MOVES default cycle with comparable average speed simulates typical urban traffic conditions with a wider range of speeds and accelerations than any modeled here. Drive-through data from the tunnel indicated that, for most drive-through cycles, vehicle speed varied little from the average; speeding up slightly until about halfway through and then slowing down slightly when exiting the tunnel. A 9-sec drive cycle (40, 40, 41, 42, 43, 42, 41, 40, 40 mph) that mimicked the average speed and pattern in 1-mph speed increments was created and used in calculating the MOVES emission factors. This drive cycle has an average speed of 41 mph, with beginning and end speeds below the average, absolute average deviation from the mean of 0.9 mph, and maximum and minimum speeds relative to the average of +2 and −1 mph, respectively.

After reviewing the initial comparisons of the modeled and measured emission factors, staff at the EPA Office Transportation and Air Quality (OTAQ) noted that the operating mode distribution of the 9-sec cycle (“speed-constructed” cycle) has a greater fraction of higher-VSP (vehicle-specific power) bins, and thus higher NO_x emission factors, than using the operating mode distribution of the combined 69 drive-through samples (“69-sample” cycle). Comparisons of the measured and modeled emission factors presented are based on calculations made by ENVIRON using the speed-constructed tunnel cycle in accordance with the study protocol requirements for a blind comparison. MOVES NO_x emission factors calculated by EPA are also presented for a cycle utilizing combined 69 drive-through samples and a flat 41-mph cycle to illustrate the model's sensitivity to the operating modes.

Ambient conditions of temperature and humidity can also affect modeled emissions, so a sample of the conditions used during the field testing was averaged to provide a typical condition. These conditions were estimated to be 88 °F and 27% relative humidity, and were used in all model calculations. As a sensitivity analysis, the exhaust and evaporative THC emission rates were also estimated using the minimum (64.7 °F) and maximum (104.4 °F) ambient temperatures during the study period. The modeled NMHC emission rates most appropriate for the conditions shown in Table 1 were used in the comparison for each measurement period (i.e., 104.4 °F for runs with ambient temperature of 101 and 102 °F; 88 °F for runs in the range of 92–95 °F; and 64.7 °F for ambient temperature of 71 and 72 °F).

The models were run with and without inspection and maintenance (I/M) program benefits included in the estimates. EMFAC used the benefits from the Smog Check program, whereas MOVES and MOBILE were run using the IM240 program credits. The IM240 program was the default I/M program for Los Angeles County in the MOVES2010a model and represents the maximum emission reduction for I/M programs. All model results shown in this paper are with I/M, unless otherwise noted. The magnitudes of the credits were determined for comparison with differences between measured and model emission factors.

Numerous other differences in modeling methods exist between the three models used, including the way in which vehicles are categorized. The EMFAC and MOBILE6.2 models segregate vehicles by body type (car, truck, or bus) and gross vehicle weight rating (GVWR). MOVES2010a identifies the vehicle types differently in order to take advantage of the classification data collected by U.S. Department of Transportation, which distinguishes vehicles only by body type and purpose (passenger cars and trucks, commercial trucks, etc.) rather than weight. Even within a particular model, the distinction between light-duty cars and light-duty trucks can be unclear because many light-duty trucks may appear to be cars, such as minivans, “crossover” sport utility vehicles, off-road station wagons, and other vehicles with sufficient ground clearance, cargo space, weight capacity, and other distinguishing characteristics from passenger cars to be defined as trucks. Classification of the vehicles by the manufacturers is also somewhat arbitrary making Department of Motor Vehicles (DMV) records, when available, unreliable for matching some types of vehicles to the model categories. In classifying the vehicles observed on the video records during tunnel measurements, each was assigned to 1 of the 13 EMFAC categories, and the cross-platform mapping system shown in Table 2 was used to translate these assignments for use with MOBILE and MOVES. As the example values in the table show, reclassification of vehicles between categories generally would not result in a significant change in the overall emission rate relative to the differences that exist between the three models.

Table 2. Vehicle categories used to assign emission factors in the three models

	EMFAC		MOBILE6		MOVES2010
Fraction of Fleet	Category	NOx (g/mi)	Category	NOx (g/mi)	Category	NOx (g/mi)
Gasoline vehicles
77.3%	Passenger car	0.07	LD vehicles (passenger cars)	0.15	Passenger car	0.09
9.9%	LD truck <3750 GVW	0.07	LD trucks 1 (0–6000 lbs. GVWR 0–3750 lbs. LVW)	0.18	Passenger truck	0.16
7.7%	LD truck <5750 GVW	0.09	LD trucks 2 (0–6001 lbs. GVWR 37515750 lbs. LVW)	0.23
0.5%	Medium-duty truck <8500 GVW	0.16	LD trucks 3 (6001–8500 lbs. GVWR 05750 lbs. ALVW)	0.49	Light commercial truck	0.24
			LD trucks 4 (6001–8500 lbs. GVWR 5751 lbs. and greater A LVW)	0.49
0.2%	Light-HD truck <10,000 GVW	0.22	HD vehicles (8501–10000 lbs. GVWR)	0.78	Single unit short-haul truck	1.95
0.0%	Light-HD truck <14,000 GVW	0.22	HD vehicles (10001–14000 lbs. GVWR)	0.80
0.0%	Medium-HD truck <33,000 GVW	0.19	HD vehicles (14001–16000 lbs. GVWR)	0.79
			HD vehicles (16001–19500 lbs. GVWR)	0.92
			HD vehicles (19501–26000 lbs. GVWR)	0.91
			HD vehicles (26001–33000 lbs. GVWR)	1.00
0.0%	Heavy-HD truck >33,000 GVW	4.37	HD vehicles (33001–60000 lbs. GVWR)	1.07	Combination short-haul truck	ND
			HD vehicles (>60000 lbs. GVWR)	ND
0.1%	Motorcycle	1.27	Motorcycles (gasoline)	0.81	Motorcycle	0.37
0.0%	School bus	ND	School buses	1.19	School bus	1.95
Diesel vehicles
0.0%	Passenger car	ND	LD vehicles (passenger cars)	0.07	Passenger car	2.20
0.0%	LD truck <3750 GVW	ND	LD trucks 1and 2 (0–6000 lbs. GVWR)	ND	Passenger truck	4.08
0.0%	LD truck <5750 GVW	ND
0.0%	Medium-duty truck <8500 GVW	ND	LD trucks 3 and 4 (6001–8500 lbs. gVWR)	0.22	Light commercial truck	4.08
0.5%	Light-HD truck <10,000 GVW	4.23	HD vehicles (8501–10000 lbs. GVWR)	2.10	Single unit short-haul truck	5.04
1.5%	Light-HD truck <14,000 GVW	4.22	HD vehicles (10001–14000 lbs. GVWR)	2.41
0.8%	Medium-HD truck <33,000 GVW	5.10	HD vehicles (14001–16000 lbs. GVWR)	2.81
			HD vehicles (16001–19500 lbs. GVWR)	3.04
			HD vehicles (19501–26000 lbs. GVWR)	3.74
			HD vehicles (26001–33000 lbs. GVWR)	4.63
0.5%	Heavy-HD truck >33,000 GVW	10.02	HD vehicles (33001–60000 lbs. GVWR)	5.25	Combination short-haul truck	9.86
			HD vehicles (>60000 lbs. GVWR)	5.90
0.0%	Urban bus	1.36	Transit and urban buses	8.08	Transit bus	8.32
0.1%	School bus	4.98	School buses	5.72	School bus	5.88
CNG vehicles
0.3%	Urban bus	ND	Transit and urban buses	ND	Transit bus	2.98
Note: NOx emission rates shown are for 2005 model year, which was the peak year of the tunnel fleet distribution.

Additionally, alternative emission rates for vehicles meeting California Low-Emission Vehicle (LEV) emission standards were used for this study. MOBILE6 and MOVES light-duty vehicle emission factors were adjusted to account for the California LEV program, which began with the 1994 model year. No adjustments for California regulations were made for vehicles older than 1994 because reliable deterioration rates do not exist for these older vehicles. Federal and California standards for heavy-duty diesel vehicles were the same for model years beginning in 1990. No adjustments were made for differences in the standards prior to 1990, as the fleet through the tunnel included very few pre-1990 diesel trucks. No other adjustments were made for California-specific motor vehicle control programs such as the Carl Moyer Program (CARB, 2012). This program was established to provide financial incentives to replace or retrofit high-polluting engines. However, this program focused on reductions of PM emissions primarily in economically disadvantaged residential areas located close to industrial areas. Effect of the program on emissions in the tunnel are likely minimal.

The modeled emission rates for each vehicle type and model year were multiplied by the number of such vehicles observed during each 3-hr measurement period. Then the resulting combined fleet emissions in grams per mile were converted to fuel-based emission rates by dividing by the fleet emissions of CO₂ plus CO and multiplying the resulting ratio by the fuel carbon mass fraction of 0.85 g per gram of fuel.

Results

This comparison is applicable to on-road vehicles in hot, stabilized mode at near-constant speeds of approximately 40 mph and excludes start emissions, and diurnal and hot-soak evaporative hydrocarbon emissions. In addition to uncertainties associated with the tunnel measurements, there are several uncertainties inherent in the emission factor modeling where modeled conditions may not match tunnel conditions. These areas of uncertainty include the vehicle drive cycle, vehicle condition, and fleet characterization. Except in the case of the MOVES model, the vehicle drive cycle used in the modeling does not represent in-use activity in a tunnel where vehicles operate at a relatively stable vehicle speed. In the case of the MOVES2010a model, the NO_x emission factors were sensitive to differences in the drive cycles or modes used to represent the driving pattern through the tunnel. Neither MOBILE nor MOVES was developed to represent the California vehicle emission control program. For the MOBILE and MOVES model results, no adjustments were made to account for the California fleet for the pre-LEV (model years prior to 1994) light-duty vehicles or other state and local emission reduction programs that might affect gasoline or diesel vehicles. Gasoline vehicles of model years prior to 1994 were estimated to account for about 25–40% of the NMHC, CO, and NO_x fleet total emissions. Although California light-duty gasoline vehicle NO_x standards were about 25% lower during most of the decade prior to 1994, the contemporary contributions of this portion of the fleet are likely related to emissions of high emitters rather than the original emission standards. University of Denver estimated from RSD measurements that more than a third of the current Sherman Way total emissions of CO and HC are contributed by less than 1% of the fleet and half of the total measured CO, HC, and NO was produced by 2.0%, 2.1%, and 5.0%, respectively, of the 13,000 measurements (Bishop et al., 2012). For these reasons, the potential differences in the contemporary in-use vehicle fleet that might be attributable to differences in the Federal and California programs are difficult to quantify.

Comparisons of measured and modeled emission factors

The measured fleet-average fuel-based emission factors (grams of pollutant per kilogram of fuel) for CO, NO_x, NMHC, EC, and TC for each of eight runs with valid data are compared in Table 3 with the corresponding model estimates using MOVES2010a, MOBILE6.2, and EMFAC2007 (estimates without I/M credits in parentheses). Uncertainty estimates of about ±30% for the measured CO, NO_x, and NMHC fuel-based emission factors are based on the propagated 1-sigma analytical errors for the background-subtracted pollutant and CO₂ concentrations. Medians of the model to measured emission factor ratios are also shown in Table 3 for weekday and weekend samples. The median, rather than the mean, values are presented in order to exclude the effect of the Sunday AM sample, which gave significantly lower emission factors for all pollutants. These low fuel-based emission rates were due to a relatively high background-subtracted CO₂ concentration on that morning, but because the individual CO₂ measurements are within the range of other samples they are still deemed to be valid.

Table 3. Comparisons of measured fleet-averaged fuel-based CO, NMHC, NO_x, EC, and TC emission factors (g/kg of fuel) and CO/NO_x and NHMC/NO_x molar ratios with corresponding model estimates using MOVES2010a, MOBILE6.2, and EMFAC2007 for each run with I/M credit (without I/M credit)

	CO	NOx	NMHC	EC	TC	CO/NOx	NMHC/NOx
Aug 21, Sat PM, 95 °F
Tunnel	23.0 ± 5.2	3.2 ± 0.8	1.59 ± 0.35	0.01 ± 0.01	0.03 ± 0.02	11.7 ± 1.6	1.48 ± 0.19
MOVES	26.9 (33.8)	4.9 (5.9)	0.98 (1.26)	0.03 (0.03)	0.08 (0.08)	9.0 (9.4)	0.6 (0.6)
MOBILE	37.9 (49.7)	4.5 (5.5)	1.48 (1.86)	0.01 (0.01)	0.04 (0.04)	13.8 (14.9)	0.9 (1.0)
EMFAC	24.0 (31.2)	2.5 (3.6)	0.91 (1.32)		0.09 (0.09)	15.6 (14.3)	1.0 (1.0)
Aug 22, Sun PM, 92 °F
Tunnel	25.4 ± 8.2	3.6 ± 1.2	1.98 ± 0.62	0.02 ± 0.01	0.07 ± 0.03	11.7 ± 1.7	1.64 ± 0.22
MOVES	27.2 (34.6)	4.4 (5.4)	1.15 (1.48)	0.01 (0.01)	0.04 (0.04)	10.2 (10.4)	0.7 (0.8)
MOBILE	39.9 (52.5)	4.2 (5.3)	1.73 (2.17)	0.00 (0.00)	0.03 (0.03)	15.5 (16.4)	1.1 (1.2)
EMFAC	24.7 (32.3)	2.0 (3.1)	1.05 (1.53)		0.08 (0.08)	19.9 (17.0)	1.4 (1.4)
Aug 24, Tue AM, 92 °F
Tunnel	16.7 ± 5.1	2.8 ± 1.1	1.40 ± 0.41	0.03 ± 0.02	0.03 ± 0.02	9.6 ± 2.8	1.35 ± 0.40
MOVES	26.2 (32.8)	5.7 (6.7)	1.22 (1.52)	0.06 (0.06)	0.13 (0.13)	7.5 (8.1)	0.6 (0.6)
MOBILE	36.5 (47.9)	4.7 (5.7)	1.63 (2.04)	0.01 (0.01)	0.05 (0.05)	12.6 (13.8)	1.0 (1.0)
EMFAC	23.0 (30.0)	2.4 (3.0)	1.02 (1.46)		0.10 (0.10)	16.0 (16.1)	1.2 (1.3)
Aug 24, Tue PM, 101 °F
Tunnel	19.1 ± 5.4	4.0 ± 1.2	2.51 ± 0.68	0.04 ± 0.02	0.06 ± 0.03	7.8 ± 1.4	1.84 ± 0.30
MOVES	30.8 (38.5)	5.1 (6.0)	1.19 (1.49)	0.06 (0.06)	0.12 (0.12)	9.9 (10.6)	0.7 (0.7)
MOBILE	36.8 (48.2)	4.8 (5.8)	1.89 (2.40)	0.01 (0.01)	0.05 (0.05)	12.5 (13.7)	1.1 (1.2)
EMFAC	23.7 (30.8)	3.4 (4.4)	1.76 (1.89)		0.10 (0.10)	11.5 (11.4)	1.5 (1.2)
Aug 25, Wed AM, 92 °F
Tunnel	18.9 ± 6.1	3.4 ± 1.2	1.35 ± 0.43	0.06 ± 0.03	0.12 ± 0.05	9.0 ± 1.9	1.05 ± 0.24
MOVES	23.6 (29.3)	5.6 (6.5)	1.17 (1.37)	0.08 (0.08)	0.16 (0.16)	7.0 (7.4)	0.6 (0.6)
MOBILE	39.0 (50.9)	5.0 (5.9)	1.69 (2.07)	0.01 (0.01)	0.05 (0.05)	12.9 (14.2)	1.0 (1.0)
EMFAC	22.5 (29.2)	3.5 (4.3)	0.91 (1.30)		0.10 (0.10)	10.6 (11.1)	0.7 (0.8)
Aug 25, Wed PM, 102 °F
Tunnel	30.4 ± 9.2	4.5 ± 1.5	3.05 ± 0.90	0.05 ± 0.03	0.11 ± 0.04	11.1 ± 1.8	2.03 ± 0.32
MOVES	27.7 (34.3)	5.3 (6.1)	1.14 (1.34)	0.08 (0.08)	0.16 (0.16)	8.6 (9.2)	0.6 (0.6)
MOBILE	38.8 (50.6)	5.3 (6.2)	2.18 (2.72)	0.01 (0.01)	0.06 (0.06)	12.1 (13.3)	1.2 (1.2)
EMFAC	22.4 (29.1)	3.7 (4.6)	1.81 (1.59)		0.10 (0.10)	9.9 (10.4)	1.4 (1.0)
Aug 28, Sat AM, 72 °F
Tunnel	25.9 ± 11.0	3.9 ± 1.7	1.09 ± 0.46	0.02 ± 0.02	0.03 ± 0.03	10.9 ± 1.9	0.71 ± 0.16
MOVES	18.1 (22.5)	4.8 (5.8)	1.18 (1.40)	0.04 (0.04)	0.10 (0.10)	6.2 (6.4)	0.7 (0.7)
MOBILE	41.7 (54.5)	4.5 (5.5)	1.63 (2.12)	0.00 (0.00)	0.04 (0.04)	15.2 (16.3)	1.0 (1.1)
EMFAC	24.0 (31.3)	2.4 (3.3)	0.81 (1.22)		0.08 (0.08)	16.5 (15.5)	0.9 (1.0)
Aug 29, Sun AM, 70 °F
Tunnel	10.7 ± 2.2	1.3 ± 0.3	0.51 ± 0.10			13.9 ± 2.4	1.11 ± 0.20
MOVES	19.7 (25.0)	4.7 (5.8)	1.21 (1.53)	0.02 (0.02)	0.07 (0.07)	6.9 (7.1)	0.7 (0.7)
MOBILE	39.8 (52.3)	4.3 (5.4)	1.48 (2.03)	0.00 (0.00)	0.03 (0.03)	15.1 (16.0)	1.0 (1.1)
EMFAC	25.0 (32.5)	2.2 (3.3)	0.92 (1.36)		0.08 (0.08)	18.6 (16.3)	1.2 (1.2)
Median ratios—Weekday
MOVES/Meas	1.41	1.46	0.67	1.56	1.77	0.78	0.40
MOBILE/Meas	1.99	1.33	0.96	0.27	0.65	1.37	0.65
EMFAC/Meas	1.22	0.84	0.69		1.29	1.33	0.74
Median ratios—Weekend
MOVES/Meas	1.12	1.38	0.85	1.89	2.52	0.67	0.55
MOBILE/Meas	1.63	1.29	1.22	0.23	1.31	1.25	0.78
EMFAC/Meas	1.01	0.70	0.66		2.60	1.42	0.97

The modeled CO emission factors from MOVES and EMFAC were both in reasonable agreement, within the measurement error, with the measured fleet-averaged emission factors. The MOBILE CO emission factors were significantly higher than the measured emission factors, with median model/tunnel ratios of 2.0 and 1.6 on weekdays and weekends, respectively. The MOVES NO_x emission factors using the speed-constructed tunnel cycle were higher than the measured emission factors for all sampling periods, with median model/tunnel ratios of 1.5 and 1.4 on weekdays and weekends, respectively. However, half of the MOVES factors were within the uncertainty of the measured factors. The sensitivity of the MOVES NO_x emission factors to vehicle operation is described in the following section. The MOBILE NO_x emission factors were also higher than the measured factors for all sampling periods with median model/tunnel ratios of 1.3 for both weekdays and weekends. EMFAC NO_x emission factors were generally lower, with median model/tunnel ratios of 0.8 for the weekday samples and 0.7 for the weekends when traffic volumes of diesel vehicles were lower. The measured NMHC emission factors for the eight sampling periods increased with ambient temperature. Dependence of measured and modeled NMHC emission factors and NMHC/NO_x ratios on ambient temperature is described in a following section.

The modeled emission factors for total particulate carbon (TC) varied among the models by about a factor of 3. Relative to MOBILE, EMFAC was twice as high and MOVES was about 3 times higher. MOVES has the greatest sensitivity to the larger fractions of diesel vehicles on weekdays (4.2%) compared with Sundays (0.9%), with emission factors that were about 3 times higher on weekdays relative to Sundays. In comparison, the weekday/Sunday ratios in TC emission factors were 1.7 for MOBILE, 1.3 for EMFAC, and about 1.1 for the measured factors. As a result, MOVES slightly underestimated measurements on Sunday and overestimated measurements by nearly a factor of 2 on weekdays. The same relative differences were obtained for comparison of MOVES EC emission factors with measurements. MOBILE TC and EC emission factors underestimated measurements during both weekdays and Sundays. TC emission factors from EMFAC were in generally in good agreement with measurements during all sampling periods.

Sensitivity of MOVES emission factors to operating cycle

Initial comparisons of modeled and measured emission factors were presented to EPA OTAQ for comment. According to EPA staff, the operating mode distribution of the speed-constructed cycle has a greater fraction of higher-VSP bins than the distribution of the 69-sample cycle, as shown in Figure 1. Figure 2 shows that by EPA's calculations, the fleet NO_x emission factors are, on average, 14% and 36% higher for the speed-constructed cycle than using operating mode distribution of all 69 drive-through samples or a flat 41-mph cycle, respectively. The median model/tunnel ratios of the NO_x emission factors are 1.5, 1.3, and 1.1 for speed-constructed, 69-sample, and flat cycles, respectively. The median ratios are within the margin of the measurement uncertainty for the 69-sample and flat cycles. All models predict some increase in evaporative hydrocarbon emissions with increasing ambient temperature; however, MOVES also includes increases in CO₂ and CO emissions related to increased use of air conditioning resulting in a decrease in fuel-based evaporative emissions with increasing temperature (see Figure 3). Because this contradicts the other models and observations of higher hydrocarbon emissions with increasing temperature, we excluded the temperature effect on CO₂ and CO in calculating the MOVES fuel-based NMHC emissions so that these factors would be directly comparable with the corresponding MOBILE and EMFAC factors.

Figure 1. Operating mode distribution by vehicle-specific power based on the speed-constructed cycle and the 69-sample cycle. Adapted from calculations provided by EPA OTAQ.

Figure 2. Comparisons of measured NO_x fuel-based emission factors by sampling period with corresponding MOVES emission factors using the “speed-contructed,” “69-sample,” and flat cycles. Adapted from calculations provided by EPA OTAQ, February 2012.

Figure 3. Temperature dependence of MOVES CO₂, CO, NO_x, and THC (exhaust) and THC (evaporative) emission factors (both distance and fuel-based) normalized to 88 °F.

Sensitivity of NMHC emission factors to ambient temperature

Both MOVES and EMFAC NMHC emission factors were in better agreement with measurements at lower temperatures. Underprediction at higher temperatures suggests that the models are not sufficiently sensitive to ambient temperature. Figure 4 shows that the measured NMHC emission factors were 2.0 and 3.5 times higher at 85–95 and 95–105 °F, respectively, compared with 65–75 °F. Figure 4 also shows the corresponding modeled emission factors for the exhaust and evaporative contributions to NMHC emissions separately. Whereas EMFAC and, to a lesser extent, MOBILE predict higher emissions with increasing temperature, the MOVES evaporative NMHC emission factors are relatively insensitive to temperature. The ratios of modeled evaporative to exhaust emission factors at 104 °F relative to 65 °F are 4.0, 1.7, and 1.1 for EMFAC, MOBILE, and MOVES, respectively. The corresponding ratio of the estimated evaporative emission factor from the tunnel measurements at ambient temperatures of 101–102 °F relative to 70–72 °F was about 6. The contributions of evaporative emissions to the measured NMHC emission factors were estimated by subtracting an average exhaust NMHC contribution of 0.4 g/kg of fuel (based on 0.5 times measured NMHC at 65 °F) from the total.

Figure 4. Sensitivity of measured NMHC emission factors to ambient temperature compared with corresponding modeled exhaust and evaporative NMHC emission factors.

The MOVES model predicts running evaporative emissions that are in reasonable agreement with the estimates from measurements at lower ambient temperatures, but underestimates the emissions at higher temperature. MOBILE predicts the highest evaporative emission factors, but has less temperature sensitivity than indicated by the measurements. The apparent temperature sensitivity observed in the tunnel measurements is best replicated by EMFAC, though it too underestimated NMHC emissions at high ambient temperatures. The measured fuel-based emission factors for species that are produced only by fuel combustion (e.g., acetylene) were insensitive to ambient temperature, whereas species that are typically enriched in gasoline vapor (e.g., n-butane) showed greater temperature dependence (Table 4). The ratios of the measured emission factors at high and low ambient temperatures were 0.9 for acetylene and 8.1 for n-butane. These results suggest that the models, especially MOVES, may not fully account for running evaporative emissions at higher ambient temperature possibly due to an underprediction, at these higher temperatures, of the vapor venting, fuel permeation, or in combination with fuel system malfunctions, such as leaks, missing gas caps, or faulty canister purge valves.

Table 4. Fuel-based emission rates (grams per kg of fuel) as a function of ambient temperature for combustion products and major components of whole gasoline

Sampling Period Temp °F	High 101–102	Med 92–95	Low 71–72	Med/Low T Ratio	Hi/Low T Ratios
Combustion Products
acetylene	0.043 ± 0.013	0.048 ± 0.016	0.051 ± 0.023	0.94	0.86
ethylene	0.021 ± 0.006	0.014 ± 0.005	0.018 ± 0.007	0.80	1.17
1,3-butadiene	0.015 ± 0.004	0.010 ± 0.003	0.013 ± 0.005	0.80	1.19
Major Components of Gasoline
n-butane	0.084 ± 0.024	0.036 ± 0.011	0.010 ± 0.005	3.42	8.12
i-butane	0.016 ± 0.005	0.009 ± 0.004	0.002 ± 0.001	5.45	9.71
i-pentane	0.726 ± 0.208	0.278 ± 0.082	0.130 ± 0.057	2.14	5.59
n-pentane	0.197 ± 0.056	0.086 ± 0.025	0.037 ± 0.016	2.34	5.35
2,2-dimethylbutane	0.052 ± 0.015	0.024 ± 0.007	0.016 ± 0.007	1.47	6.47
2,3-dimethylbutane	0.044 ± 0.013	0.018 ± 0.005	0.007 ± 0.002	2.58	3.18
2-methylpentane	0.078 ± 0.022	0.039 ± 0.012	0.028 ± 0.012	1.42	3.09
3-methylpentane	0.134 ± 0.041	0.067 ± 0.022	0.043 ± 0.019	1.55	2.84
n-hexane	0.054 ± 0.016	0.029 ± 0.009	0.019 ± 0.008	1.57	2.87
2,3-dimethylpentane	0.058 ± 0.017	0.034 ± 0.010	0.023 ± 0.009	1.50	2.56
3-methylhexane	0.040 ± 0.012	0.024 ± 0.007	0.018 ± 0.008	1.31	2.17
2,2,4-tri methyl pentane	0.010 ± 0.003	0.006 ± 0.002	0.005 ± 0.002	1.22	1.98
n-heptane	0.021 ± 0.006	0.014 ± 0.004	0.006 ± 0.002	2.46	3.57
2,3,4-tri methyl pentane	0.023 ± 0.007	0.017 ± 0.005	0.014 ± 0.006	1.24	1.67
toluene	0.114 ± 0.033	0.094 ± 0.028	0.071 ± 0.031	1.33	1.60
ethylbenzene	0.018 ± 0.005	0.015 ± 0.005	0.012 ± 0.005	1.30	1.58
m&p-xylenes	0.066 ± 0.019	0.056 ± 0.016	0.043 ± 0.019	1.28	1.52
o-xylene	0.024 ± 0.007	0.020 ± 0.006	0.015 ± 0.007	1.33	1.58
NMHC
Measured - Tunnel	2.78 ± 0.80	1.58 ± 0.46	0.80 ± 0.33	1.97	3.45
MOVES	1.17	1.13	1.20	0.94	0.98
MCBILE6	2.03	1.63	1.56	1.05	1.30
EMFAC	1.78	0.97	0.86	1.13	2.06
NHMC/NOx
Measured - Tunnel	1.93 ± 0.31	1.38± 0.28	0.91 ± 0.18	1.52	2.13
MOVES	0.63	0.72	0.71	1.01	0.89
MCBILE6	1.13	1.00	0.99	1.01	1.14
EMFAC	1.41	1.10	1.06	1.04	1.34
Note: Measured and model total NMHC are also shown.

NMHC and NO_x are the key precursors of ozone, and the ratios of NMHC to NO_x affect the rate and efficiency of ozone formation in the atmosphere. The comparisons of the measured NMHC/NO_x ratios with corresponding modeled ratios in Figure 5 illustrate the greater underestimation with increasing ambient temperature. Whereas the predicted NMHC/NO_x ratios are in good agreement with the measured ratios during cooler sampling periods, the measured NMHC/NO_x ratios are 3.1, 1.7, and 1.4 times higher during high-temperature periods than predicted by MOVES, MOBILE, and EMFAC, respectively.

Figure 5. Comparison of measured and modeled NMHC/NO_x molar ratios for three temperature ranges.

Model estimates of relative contributions of gasoline and diesel vehicles to fleet emissions

Separate emission factors can be estimated for gasoline and diesel vehicles from measurements in tunnels with mixed traffic by extrapolating the linear regressions of fleet-averaged emission factors with fractions of diesel vehicles to each end of the abscissa. This was not done in this study due to the consistently low fractions of diesel vehicles (from 0.9% on Sundays to 4.2% on weekdays) and the resulting large uncertainties in the extrapolated emission factors. Sensitivity of the fleet-averaged emission factors from the tunnel measurements to variations in fractions of diesel and gasoline vehicles was examined by calculating modeled fuel-based emission factors separately for the gasoline and diesel vehicles that passed through the tunnel during each sampling period. The resulting average modeled NMHC, NO_x, and total carbon emission factors for all vehicles, only gasoline vehicles, and only diesel vehicles are compared in Figure 6 with the measured mean weekday and Sunday fleet-averaged emission factors. Variations in the fleet fractions of diesel vehicles have negligible effect on fleet-averaged NMHC emission factors. Most of the differences are due to the higher average ambient temperatures on weekdays relative to Sundays during the study period.

Figure 6. Comparisons of measured fleet-averaged NMHC, NO_x, and TC emission factors with modeled emission factors for all vehicles, only gasoline vehicles, and only diesel vehicles.

Unlike NMHC and CO, the measured fleet-averaged emission factors for NO_x and TC are expected to be higher on weekdays due to higher fractions of diesel vehicles, which have much higher NO_x and TC emission rates than gasoline vehicles. Therefore, the modeled NO_x and TC emission factors for gasoline vehicles should be less than or equal to the measured fleet-averaged emission factors. This is the case, within the uncertainties, for EMFAC. This is not the case for MOVES and MOBILE models, which predict higher gasoline-only emission factors than the fleet-averaged tunnel measurements and therefore result in higher modeled versus measured fleet-averaged NO_x and TC emission factors. Figure 7 shows how the three models distribute emissions by model year within the two fuel categories for a typical weekday fleet observed at the Van Nuys tunnel. Although the overall patterns are similar for all models, there are substantial differences in the distribution for NMHC from gasoline vehicles in MOBILE and for NO_x from gas vehicles in EMFAC. Also, MOVES attributes much more NMHC to diesel vehicles in the 2000–2006 model years than the other models.

Figure 7. Gasoline and diesel fleet-averaged NMHC and NO_x emission rates (g/mi) by model year for MOVES2010a, MOBILE6.2, and EMFAC2007. Fleet distribution is from the Tuesday AM tunnel run.

The modeled emission factors and observed traffic volumes by vehicle type were used to estimate the relative contributions of gasoline and diesel vehicles to total fleet CO, NO_x, and TC emissions, as shown in Figure 8. Although gasoline vehicles are the dominant source of CO with all three models, large variations exist among the models for relative contributions of diesel and gasoline vehicles to total NO_x and TC emissions. With an average fleet fraction of 4.2% on weekdays, the diesel vehicle contributions to total NO_x emissions were 50% using EMFAC, 30% for MOVES, and 20% for MOBILE and contributions to total carbon were 75% using MOVES and about 50% for both EMFAC and MOBILE. These comparisons illustrate the varying contributions of diesel and gasoline vehicles to total NO_x and PM emissions that would be estimated using EMFAC or MOVES, and the potential changes from a transition from MOBILE to MOVES.

Figure 8. Relative contributions of gasoline and diesel vehicles to modeled fleet-averaged CO, NO_x, and total particulate carbon emissions for average weekday or Sunday runs.

Comparisons of RSD and model emission factors

RSD measurements were included in the study to examine the distribution of emissions in the local fleet of light-duty vehicles (Bishop et al., 2012). As configured for this study, the RSD could not detect elevated exhaust plumes from trucks and buses. Therefore, comparisons of RSD and tunnel measurements are applicable to only light-duty vehicles and tailpipe emissions, which would exclude evaporative emissions. Considering these limitations, the most appropriate comparison of RSD and tunnel measurements is for CO, and possibly for NO_x on weekends when there were minimal heavy-duty vehicles. Table 5 shows the average remote sensing measurements compared with the mean (± standard error) of the weekend tunnel measurements with and without the Sunday AM sample. Despite differences in location and time period of the measurements, fleet-averaged fuel-based CO and NO_x emission factors from the tunnel during weekends and RSD measurements agreed reasonably well for both CO and NO_x.

Table 5. Comparisons of average RSD and modeled fuel-based CO, NO_x, and THC emission factors

	CO	NOx	THC (exhaust)
Tunnel, weekend	21.3 ± 3.7	3.0 ± 0.6
Tunnel, weekend—Sun AM	24.8 ± 0.9	3.6 ± 0.2
Remote sensing	19.7 ± 1.3	4.0 ± 0.1	2.3 ± 0.3
MOVES	27.4	4.4	0.8
MOBILE	40.8	4.2	1.1
EMFAC	25.0	2.0	0.6

The modeled emissions for approximately 13,000 light-duty gasoline vehicles included in the remote sensing tests were averaged and were compared with the mean RSD fuel-based CO, THC, and NO_x emission factors in Table 5. MOVES and EMFAC gave average CO emission factors that were 1.4 and 1.3 times higher than the averages of the 13,000 RSD measurements, respectively, whereas the average MOBILE CO emission factor was more than double the RSD average. The average MOVES and MOBILE NO_x emission factors were only slightly higher than RSD measurements. However, the EMFAC NO_x emission factors were about half the average RSD factors, which were consistent with underprediction by EMFAC for weekend periods relative to the measured NO_x emission factors from the tunnel experiment. The modeled THC emission factors were about a third of the average RSD emission factor for both MOVES and EMFAC and about half for MOBILE.

Discussion and Conclusions

As motor vehicles are major sources of both VOC and NO_x emissions in urban area, changes in ozone photochemistry, historic trends in ambient ozone levels, and the magnitude and spatial extent of the weekend ozone effect have been closely linked to changes in vehicle emissions (Chinkin et al., 2003; Fujita et al., 2003a, 2003b). It is well established that mobile source emissions in the South Coast Air Basin (SoCAB) for carbon monoxide and hydrocarbons were underestimated in past emission inventories by as much as a factor of 2–3 relative to NO_x (Fujita et al., 1992; Harley et al., 1993; Ingall et al., 1989; Pierson et al., 1990). During the 1987 Southern California Air Quality Study (SCAQS), the on-road emission rates measured in the Van Nuys tunnel were more than 2 times larger than those calculated by EMFAC (version 7E at the time) for hydrocarbons and CO (Ingalls et al., 1989). Additionally, the ambient VOC/NO_x ratios measured during SCAQS (∼8–10 in ppb C to ppb NO_x) were about 2–2.5 times higher than the corresponding emission inventory ratios (∼4) (Fujita et al., 1992). The SCAQS database was used by the California Air Resources Board, the South Coast Air Quality Management District, and Carnegie Melon University/California Institute of Technology for air quality model evaluations and all obtained predicted ozone values that were much lower than observed (Chico et al., 1993; Harley et al., 1993; Wagner and Wheeler, 1993). Because ozone formation is most efficient at VOC/NO_x ratios near 10, the actual rate of ozone formation was faster than expected from the base model results. Model performance was greatly improved in model sensitivity tests by increasing the total on-road motor vehicle NMHC emissions by a factor of 2.5 over the official inventory. In this study, we derived fleet-averaged fuel-based emission factors from on-road emission measurements at the same tunnel used in the 1987 SCAQS. The 2010 Van Nuys tunnel study is part of an evaluation of the emission inventory and update of the 1987 “top-down” ambient versus inventory reconciliation analysis for the SoCAB.

A significant finding of the 2010 Van Nuys tunnel study was that measured NMHC fuel-based emission factors were about 3.5 times higher during high-temperature periods (101–102 °F) than cool periods (71–72 °F). The increased emissions during hot periods were attributed to light hydrocarbons that are associated with headspace evaporative emissions. These results are generally consistent with an ambient source apportionment study that estimated a 6.5% ± 2.5% increase in the contributions of evaporative emissions from motor vehicles per degree Celsius increase in maximum temperature (Rubin et al., 2006). Although the NMHC emission rates predicted by all three models were in good agreement with measurements during cool periods, the running evaporative emissions for all models exhibited insufficient sensitivity to temperature during hot periods, especially MOVES. The measured NMHC/NO_x ratios were 3.1, 1.7, and 1.4 times higher than predicted by MOVES, MOBILE, and EMFAC, respectively, during hot periods. These results suggest that there is an overall increase in motor vehicle NMHC emissions on hot days that is not fully accounted for by the emission models. Hot temperatures and concomitant higher ratios of NMHC emissions relative to NO_x both contribute to more rapid and efficient formation of ozone.

Another significant finding relates to the sensitivity of the MOVES to vehicle operating modes, which is an essential design element of the model. The speed-constructed 9-sec drive cycle used by ENVIRON to estimate the MOVES emission factors for our blind comparisons with the measured emission factors yields about 14% and 36% higher NO_x emission factors than the vehicle operating mode distributions of the combined 69 drive-through samples and 41-mph flat cycle, respectively. Although the MOVES NO_x emission factors reported here were generally higher than the measured factors, most differences were not significant considering this factor. The variations related to sensitivity of MOVES to operating mode increases the uncertainty of the comparisons with the measured emission factors and illustrate the importance of selecting the appropriate operating modes for project-level analysis.

Irrespective of the model with measurement comparisons, we observed large variations among the three models in the predicted emission factors and relative contributions of diesel and gasoline vehicles to total NO_x and particulate carbon (TC) emissions in the tunnel. During weekday, diesel trucks accounted for 33% of the total NO_x emissions in the tunnel according to MOVES and 50% by EMFAC. Contributions of diesel trucks to total carbon were 75% by MOVES and 30% by EMFAC. Although the MOVES NO_x emission factors were higher than EMFAC for gasoline vehicles (4.0 vs. 1.8 g/kg of fuel), they were lower for diesel vehicles (15.6 vs. 18.0 g/kg of fuel). The MOVES TC emission factors were 0.03 and 0.95 g/kg for gasoline and diesel, respectively, whereas the corresponding EMFAC TC emission factors were 0.08 and 0.33 g/kg. These differences may be partly related to the aforementioned differences in the federal and California emission standards and emission control programs. Another possible explanation may be related to the cycle-average approach used in EMFAC versus the MOVES modal modeling approach. Vehicle emission tests have shown that a large fraction of the PM emissions from normal emitters are associated with hard acceleration events in the vehicle test cycle (Fujita et al., 2007). Measurements during the Kansas City Vehicle Emissions Characterization Study showed that the hard acceleration event that occurs between 840 and 880 sec of the unified driving cycle accounted for about 40% of the total hot running PM emission (Lindhjem et al., 2009) and that this fraction was similar for all model years newer than 1980. The modal approach should predict lower PM emissions than the cycle-average approach for the relatively stable driving pattern in the tunnel.

Acknowledgments

This project was funded by the U.S. Department of Energy Office of Vehicle Technologies (Dr. James Eberhardt, Chief Scientist) through the National Renewable Energy Laboratory. We thank the staff of the Air Quality and Modeling Center of the U.S. EPA Office of Transportation and Air Quality for their review of the draft manuscript and for providing additional model calculations. The authors thank the following DRI personnel for their assistance with sample analysis: Anna Cunningham and Mark McDaniel for organic speciation analysis, and Steven Kohl, Ed Hackett, and Brenda Cristani for analysis of inorganic species. We also acknowledge the assistance of Andy Chew of Andy's Automotive for processing the video recordings of traffic through the Van Nuys tunnel.